Xiantao Zhang

Wenwen Zhuang, Xin Huang, Xiantao Zhang et al.

Multimodal Large Language Models (MLLMs) excel in solving text-based mathematical problems, but they struggle with mathematical diagrams since they are primarily trained on natural scene images. For humans, visual aids generally enhance problem-solving, but MLLMs perform worse as information shifts from textual to visual modality. This decline is mainly due to their shortcomings in aligning images and text. To tackle aforementioned challenges, we propose Math-PUMA, a methodology focused on Progressive Upward Multimodal Alignment. This approach is designed to improve the mathematical reasoning skills of MLLMs through a three-stage training process, with the second stage being the critical alignment stage. We first enhance the language model's mathematical reasoning capabilities with extensive set of textual mathematical problems. We then construct a multimodal dataset with varying degrees of textual and visual information, creating data pairs by presenting each problem in at least two forms. By leveraging the Kullback-Leibler (KL) divergence of next-token prediction distributions to align visual and textual modalities, consistent problem-solving abilities are ensured. Finally, we utilize multimodal instruction tuning for MLLMs with high-quality multimodal data. Experimental results on multiple mathematical reasoning benchmarks demonstrate that the MLLMs trained with Math-PUMA surpass most open-source MLLMs. Our approach effectively narrows the performance gap for problems presented in different modalities. The code and data are available at: \url{https://github.com/wwzhuang01/Math-PUMA}.

Yun Piao, Hongbo Min, Hang Su et al.

The rapid advancement of Large Language Models (LLMs) is catalyzing a shift towards autonomous AI Agents capable of executing complex, multi-step tasks. However, these agents remain brittle when faced with real-world exceptions, making Human-in-the-Loop (HITL) supervision essential for mission-critical applications. In this paper, we present AgentBay, a novel sandbox service designed from the ground up for hybrid interaction. AgentBay provides secure, isolated execution environments spanning Windows, Linux, Android, Web Browsers, and Code interpreters. Its core contribution is a unified session accessible via a hybrid control interface: An AI agent can interact programmatically via mainstream interfaces (MCP, Open Source SDK), while a human operator can, at any moment, seamlessly take over full manual control. This seamless intervention is enabled by Adaptive Streaming Protocol (ASP). Unlike traditional VNC/RDP, ASP is specifically engineered for this hybrid use case, delivering an ultra-low-latency, smoother user experience that remains resilient even in weak network environments. It achieves this by dynamically blending command-based and video-based streaming, adapting its encoding strategy based on network conditions and the current controller (AI or human). Our evaluation demonstrates strong results in security, performance, and task completion rates. In a benchmark of complex tasks, the AgentBay (Agent + Human) model achieved more than 48% success rate improvement. Furthermore, our ASP protocol reduces bandwidth consumption by up to 50% compared to standard RDP, and in end-to-end latency with around 5% reduction, especially under poor network conditions. We posit that AgentBay provides a foundational primitive for building the next generation of reliable, human-supervised autonomous systems.

Xianhang Ye, Yiqing Li, Wei Dai et al.

Existing GUI grounding methods often struggle with fine-grained localization in high-resolution screenshots. To address this, we propose GUI-ARP, a novel framework that enables adaptive multi-stage inference. Equipped with the proposed Adaptive Region Perception (ARP) and Adaptive Stage Controlling (ASC), GUI-ARP dynamically exploits visual attention for cropping task-relevant regions and adapts its inference strategy, performing a single-stage inference for simple cases and a multi-stage analysis for more complex scenarios. This is achieved through a two-phase training pipeline that integrates supervised fine-tuning with reinforcement fine-tuning based on Group Relative Policy Optimization (GRPO). Extensive experiments demonstrate that the proposed GUI-ARP achieves state-of-the-art performance on challenging GUI grounding benchmarks, with a 7B model reaching 60.8% accuracy on ScreenSpot-Pro and 30.9% on UI-Vision benchmark. Notably, GUI-ARP-7B demonstrates strong competitiveness against open-source 72B models (UI-TARS-72B at 38.1%) and proprietary models.

Zhengan Yan, Shikang Zheng, Haoran Qin et al.

Diffusion-based image editing offers strong semantic controllability, but remains computationally expensive due to iterative high-resolution denoising over all spatial tokens. Dynamic-resolution sampling reduces this cost by performing early steps at reduced resolution. However, existing approaches prioritize upsampling using low-level heuristics such as edge detection or channel variance, which are weakly aligned with editing semantics and may lead to structural inconsistency. Moreover, spatial regions are often upsampled without verifying whether semantic modification is actually required, resulting in redundant high-resolution computation and accumulated errors. Therefore, we propose SpecEdit, a training-free dynamic-resolution framework tailored for diffusion-based image editing. SpecEdit follows a draft-and-verify scheme: a low-resolution draft first estimates the semantic outcome, after which token-level discrepancies are used to identify edit-relevant tokens for high-resolution denoising, while the remaining tokens stay at a coarse resolution. Experiments on Qwen-Image-Edit and FLUX.1-Kontext-dev demonstrate up to 10x and 7x acceleration, while maintaining strong quality. SpecEdit is complementary to step distillation and other acceleration techniques, achieving up to 13x speedup when combined with existing methods. Our code is in supplementary material and will be released on GitHub.

Xiantao Zhang

Multimodal Large Language Models (MLLMs) hold immense promise as assistive technologies for the blind and visually impaired (BVI) community. However, we identify a critical failure mode that undermines their trustworthiness in real-world applications. We introduce the Escalator Problem -- the inability of state-of-the-art models to perceive an escalator's direction of travel -- as a canonical example of a deeper limitation we term Implicit Motion Blindness. This blindness stems from the dominant frame-sampling paradigm in video understanding, which, by treating videos as discrete sequences of static images, fundamentally struggles to perceive continuous, low-signal motion. As a position paper, our contribution is not a new model but rather to: (I) formally articulate this blind spot, (II) analyze its implications for user trust, and (III) issue a call to action. We advocate for a paradigm shift from purely semantic recognition towards robust physical perception and urge the development of new, human-centered benchmarks that prioritize safety, reliability, and the genuine needs of users in dynamic environments.

Xiantao Zhang

Visually rich documents (VRDs) challenge retrieval-augmented generation (RAG) with layout-dependent semantics, brittle OCR, and evidence spread across complex figures and structured tables. This survey examines how Multimodal Large Language Models (MLLMs) are being used to make VRD retrieval practical for RAG. We organize the literature into three roles: Modality-Unifying Captioners, Multimodal Embedders, and End-to-End Representers. We compare these roles along retrieval granularity, information fidelity, latency and index size, and compatibility with reranking and grounding. We also outline key trade-offs and offer some practical guidance on when to favor each role. Finally, we identify promising directions for future research, including adaptive retrieval units, model size reduction, and the development of evaluation methods.

Ziyuan Chen, Zhenghui Zhao, Zhangye Han et al.

With the rapid advancement of large language models and vision-language models, employing large models as Web Agents has become essential for automated web interaction. However, training Web Agents with reinforcement learning faces critical challenges including credit assignment misallocation, prohibitively high annotation costs, and reward sparsity. To address these issues, we propose Tree-Guided Preference Optimization (TGPO), an offline reinforcement learning framework that proposes a tree-structured trajectory representation merging semantically identical states across trajectories to eliminate label conflicts. Our framework incorporates a Process Reward Model that automatically generates fine-grained rewards through subgoal progress, redundancy detection, and action verification. Additionally, a dynamic weighting mechanism prioritizes high-impact decision points during training. Experiments on Online-Mind2Web and our self-constructed C-WebShop datasets demonstrate that TGPO significantly outperforms existing methods, achieving higher success rates with fewer redundant steps.

Xiaoxian Guo, Xiantao Zhang, Xinliang Tian et al.

The real-time motion prediction of a floating offshore platform refers to forecasting its motions in the following one- or two-wave cycles, which helps improve the performance of a motion compensation system and provides useful early warning information. In this study, we extend a deep learning (DL) model, which could predict the heave and surge motions of a floating semi-submersible 20 to 50 seconds ahead with good accuracy, to quantify its uncertainty of the predictive time series with the help of the dropout technique. By repeating the inference several times, it is found that the collection of the predictive time series is a Gaussian process (GP). The DL model with dropout learned a kernel inside, and the learning procedure was similar to GP regression. Adding noise into training data could help the model to learn more robust features from the training data, thereby leading to a better performance on test data with a wide noise level range. This study extends the understanding of the DL model to predict the wave excited motions of an offshore platform.

Xiaoxian Guo, Xiantao Zhang, Xinliang Tian et al.

Real-time motion prediction of a vessel or a floating platform can help to improve the performance of motion compensation systems. It can also provide useful early-warning information for offshore operations that are critical with regard to motion. In this study, a long short-term memory (LSTM) -based machine learning model was developed to predict heave and surge motions of a semi-submersible. The training and test data came from a model test carried out in the deep-water ocean basin, at Shanghai Jiao Tong University, China. The motion and measured waves were fed into LSTM cells and then went through serval fully connected (FC) layers to obtain the prediction. With the help of measured waves, the prediction extended 46.5 s into future with an average accuracy close to 90%. Using a noise-extended dataset, the trained model effectively worked with a noise level up to 0.8. As a further step, the model could predict motions only based on the motion itself. Based on sensitive studies on the architectures of the model, guidelines for the construction of the machine learning model are proposed. The proposed LSTM model shows a strong ability to predict vessel wave-excited motions.